AIX status

[Benjamin Kosnik]

> Phil> The thread targets aren't for "posix under machine X,"
> Phil> they're for entirely different kinds of threads.  If you want to use posix
> Phil> threads, then we need to make --enable-threads=posix work properly, rather
> Phil> than introducing a "synonym wrapper".  IMO.
> 
> 	I would be happy to collaborate with anyone to allow
> --enable-threads=posix to "work properly" on AIX given AIX's requirements.

I appreicate this spirt of cooperation.....

> Some of the previous comments lead me to believe that "posix" support
> should not be cluttered with AIX's needs, so I was trying to keep the
> changes localized to configure and not affecting common files and
> configure case.

Yep. I think you understand the issues around the reluctance for getting 
your current patches in.

> 	Can the pthreads case for configure include a special set of AIX
> tests for when to use threads-posix.h and threads-no.h?

Yes. 

> 	Can threads-no.h and threads-posix.h be merged and wrapped by
> _GLIBCPP_USE_THREADS after including c++config.h to choose which set of
> definitions?

Errrrr. No. Why do you want to do this?

> 	AIX needs to include pthread.h and define the v3/libio threads
> symbols when the -pthread compile/link option is present and not reference
> those symbols when the -pthread compile/link option is absent.  AIX does
> not (cannot) provide weak, no-op versions of the symbols when the option
> is absent.

Hmm. Could you put dummy functions (stubs) in instead of weak linking?


-benjamin